Project AURIONAI Use-Cases Research & Innovation
Sea AI Lab
vision

MetaFormer (PoolFormer, ConvFormer, CAFormer)

Sea AI Lab family that keeps the transformer block layout but swaps the token mixer, starting with pooling in PoolFormer.

Sea AI LabSea AI LabSingaporeNational University of SingaporeNational University of SingaporeSingapore
Figure from the MetaFormer (PoolFormer, ConvFormer, CAFormer) paper: The overall framework of PoolFormer.
Levelmodel
Classvision
Parameters11.92M
Contextnot in the notes
Released22 Nov 2021
Opennessopen weights permissive
Runs onnot in the notes
Statusactive

R&D Topics

Image classification backbone architectures (CNN and vision transformer)produced by

Identity

Developer
Sea AI Lab, National University of Singapore
Family
not in the notes
Released
22 Nov 2021 sourcesannouncement page
Weights published
2 Mar 2022
Record checked
12 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
other
Total parameters
11.92M sources
Active parameters
–
Checkpoint parameters
–
Note
PoolFormer v1 and v2, ConvFormer and CAFormer (arXiv 2210.13452).
Context length
not in the notes
Input
image
Output
labels and boxes
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights permissive
Licence
Apache License 2.0 sources
Access routes
open weights download, cloud marketplace
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Benchmark Results

The notes record no benchmark result for this model.

Lineage

The notes record no lineage for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. arxiv.org/abs/2111.11418 https://arxiv.org/abs/2111.11418
  2. huggingface.co/sail/poolformer_s12 https://huggingface.co/sail/poolformer_s12
  3. ai.azure.com/catalog/models?task=image-classification https://ai.azure.com/catalog/models?task=image-classification
  4. github.com/open-mmlab/mmsegmentation/tree/main/configs/p... https://github.com/open-mmlab/mmsegmentation/tree/main/configs/poolformer
  5. github.com/huggingface/pytorch-image-models/blob/main/RE... https://github.com/huggingface/pytorch-image-models/blob/main/README.md
  6. arxiv.org/abs/2210.13452 https://arxiv.org/abs/2210.13452
  7. arxiv.org/html/2111.11418 https://arxiv.org/html/2111.11418

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "MetaFormer (PoolFormer, ConvFormer, CAFormer)", https://aurion.semai.id/models/metaformer/, records of October 2026.